A new open-source software package is sending shockwaves through the tech world and raising serious concerns about the future of digital identity. The tool, called Deep-Live-Cam, swaps a person’s face onto a live video stream in real-time, and it’s honestly scary as hell.
Deep-Live-Cam leverages a combination of existing AI models to achieve its impressive results. It starts by detecting faces in both a source image and a live video feed. Then, using a model called “inswapper,” it swaps the faces. Finally, another model, GFPGAN, enhances the swapped face to make it look more realistic. The underlying technology, trained on vast datasets of facial images, has learned to understand facial structures and dynamics, allowing it to create highly convincing deepfakes.
X user João Fiadeiro demonstrated the tool, replacing his face with J.D. Vance, Hugh Grant, Mark Zuckerberg, and George Clooney:
Real-time deepfakes are here. I used the repo below to swap my face directly in my webcam feed as JD Vance, Hugh Grant, Zuckerburg, and George Clooney.We know people have been using this kind of tech for malicious purposes… Don’t trust your eyes, don’t trust your ears. pic.twitter.com/iWvgtwtXLD— joao (@jay_wooow) August 9, 2024
Possible concerns
While the initial applications of Deep-Live-Cam might seem harmless – think of fun social media filters or impersonating celebrities – the potential for misuse is alarming. ” Weird how all the major innovations coming out of tech lately are under the Fraud skill tree,” illustrator Corey Brickley wrote on Twitter. “Remember to establish code words with your parents everyone.”
Weird how all the major innovations coming out of tech lately are under the Fraud skill tree— Corey Brickley Illustration. Justice for Palestine (@CoreyBrickley) August 13, 2024
Imagine a scenario where a scammer uses a deepfake to impersonate a trusted colleague or family member, convincing you to transfer money or share sensitive information. Or consider the potential for deepfakes to be used to spread misinformation or manipulate public opinion. The possibilities are endless and equally terrifying.
The concerns extend beyond facial deepfakes. Voice cloning and video image-to-image AI synthesis are also rapidly evolving. While not yet at the same level of sophistication as face-swapping technology, these tools are on a trajectory to become more accessible and realistic in the near future.
So, what can be done about this? The first thing that comes to mind is the development of countermeasures to detect deepfakes. Researchers and tech companies are working on AI-powered tools to identify manipulated content. However, it’s a constant arms race, as deepfake technology continues to advance.
Additionally, but not less important, raising public awareness about deepfakes is essential. People need to be educated about the risks and how to spot potential deepfakes and AI-generated images. Critical thinking and verification of information will become increasingly important in the digital age. Although, when I see people on Facebook who believe mandala-colored parrots are real… I’m not particularly optimistic about the latter.
[via PetaPixel, Ars Technica]
